iNotbot

Humanity is losing the battle

I conducted a little experiment over the weekend.

I should start by explaining:

• I have a very small following on X, and I suspect 95% of them are bots and/or dead accounts.

• My X account exists so that someone doesn’t create an X account to masquerade as me, and research stuff like this post.

• I’ve think I’ve received maybe 5 direct messages that weren’t bots / scammers since I created my X account in 2009.

I started posting more frequently on X a couple days ago.

I also started replying to more posts in my “For You” feed.

My observations

1. Despite having the lovely paid “Blue Check”, I get almost zero views organically via the X Algo

2. I also noticed that 90% of my replies have very few views

3. The interesting thing about the replies is that suddenly I’m picking up followers left and right, but I’m not getting replies or likes to my replies

4. In some cases I likely replied to an aggregator / harvester bot account which explains the strange lack of views despite picking up followers

5. After about 24 hours I started getting Direct Messages from some new followers. 90% of them were blatantly obviously bots. The “Hi, how are you today?” opener combined with a bunch of random numbers in the username is all you need.

6. But a couple made it through my sleep-deprived brain this morning as being “slight chance of not being a bot”.

My Hypothesis

We aren’t nearly as close to AGI as all of the so called “experts” would have you believe.

Yes, the models are getting better at a lot of things…but a rational thinking human can still detect them just by talking to them.

But more seriously I think this is one of the symptoms of the Dead Internet theory.

Dead Internet Theory

The Dead Internet Theory started out as a “conspiracy theory” in 2016, but it is beginning to look like truth not conspiracy.

The theory alleges that starting in 2016, the majority of traffic on the internet was bots and AI generated content, and obviously growing. And this was before the revolution in AI.

Essentially, the internet will become worthless to humans and we will end up with AI bots arguing with each other all day long.

That will be an interesting thing to watch…how do social media companies make money with no humans in the loop?

Do the bots start paying each other? Do the companies charge the bots?

It sounds ludicrous, but the bots are good at scamming.

Imagine a bot that could self-fund through scamming, store its money in crypto tokens, setup new X accounts, operate autonomously, but when a victim gets suspicious hand the conversation over to a human scammer…or to a better more expensive LLM.

Imagine as AGI approaches how much worse this gets.

Experiment

I started chatting with the potential “not 100% sure it is a bot” bots to see what would happen.

Within a couple of back and forth chats it was completely obvious that both were bots.

Example #1 - Jessica from Miami

Referring to my profile pic, “Jessica” says:

Ok, so off to a possible “notbot” start.

But my profile does say Colorado on it, so a bot could scrape my profile, run an AI image search and possibly formulate that question.

Or a human writes the first question and then hands it off to a bot to monitor / respond to more likely.

Can Jessica make it to round 2?

Oof. Not looking so good for team NotBot here. How did we jump to "champagne glasses”, mentorship and I’m a professional?

Let’s see round 3:

Hmm.

• Laughing seems human

• Not quite sure how 50 and 35 are about the same age (although the bot is clearly trying to flirt at times later on)

• Do female human accounts get male flirty bots? I assume so but don’t know. Do gay men get gay bots?

• Second time we are clinking glasses in 3 rounds. Feels like a bot trick to seem more human by injecting emoji’s.

What does ZeroGPT say?

So, now for my favorite moment:

HAHAHAHAHA!!!!!!!!!!

Nope. It never snows in the mountains where people go to ski in the winter.

A bot from Singapore working in Miami should be just fine solo camping in a blizzard.

And what exactly is a professional local? I’m an amateur at best.

Despite that I continued chatting for awhile to see if the bot would ever try to scam me.

Eventually it got bored with me and sent this:

Those who know me might strongly disagree with Jessica’s assessment of me. But the bot finally took action. Trying to move me off X and onto WhatsApp for whatever reason.

So I ignored it and then I get another attempt to move me to WhatsApp with a little more aggressive attitude:

Example #2 - Maria from LA

This one had me fooled at first.

Although I’m a little fuzzy this morning, as I enjoyed one too many THC beverages last night.

Can dumb not fully awake Brian detect this bot?

I should explain that one of my replies as part of my experiment was replying to the “@Puppieslover” account” with a picture of one of our dogs.

NotBot Maria starts off like this:

So NotBot Maria saw or scraped my reply, managed to comment on what made the dog cute and asked for his name. Could still be a human.

And Maria said “tweets”, which in my brain is what I still call posts on X, so it felt human to me.

I didn’t actually see this message come in until a day later when “Maria” decided to ping me again despite not responding to the original question about the dog’s name.

Well damn, I don’t have a new friend in LA. Even dumb me recognizes this as typical LLM generated phrases.

I have the sad realization that 100% of my new followers on X are bots.

ZeroGPT’s verdict?

Much like before, I continued chatting with “Maria” for awhile. Actually, I’m still talking to “Maria”.

She/It has yet to ask me for anything.

Maria is definitely running better code than Jessica.

At one point, Maria sends me a picture of a Bichon Frise, tells me her name is Daisy, says Daisy looks funny when she lies down, etc.

At this point I’m having a little doubt creep in, maybe ZeroGPT was wrong. Or maybe a human stepped in to get my attention before handing me back to MariaBot.

I reverse image search the Bichon Frise and receive zero hits. But that doesn’t really mean anything since AI generates images that are quite good at times.

A Bichon Frise is consistent with my mental stereotype of LA too.

So I continue chatting with “Maria” for a bit since the F1 race was boring with Max just circling out in front by himself.

Maria starts probing for information.

Asking me what I do for work, phrased almost identically to the way Jessica asked the question at one point.

The final blow to Maria’s humanity tells you how sloppy Maria’s coders are.

You literally have my name in the chat window.

You are following my account that uses my real name.

Now after chatting for an hour you suddenly don’t know my name?

Takeaways

• Think critically about all communications with strangers. I know, it is obvious. But you have to play defense. You have to assume the entity you are interacting with is not human.

• When in doubt, try an AI detector like ZeroGPT.com

• What’s the point of engaging with X if you are just acquiring an army of bot followers?

I believe the new LLM based AI offerings are rapidly accelerating this.

I intend to do multiple follow-up posts on this topic.

Namaste.